AstroLLaMA-2-70B
收藏资源简介:
AstroLLaMA-2-70B数据集是由伊利诺伊大学厄巴纳-香槟分校的研究团队创建的,旨在评估天文领域专用大型语言模型(LLMs)的性能。该数据集包含4425个多选题(MCQs),涵盖广泛的天文学主题和概念。数据集的创建过程包括从arXiv的天文物理学类别中提取摘要、引言和结论部分,并通过光学字符识别(OCR)技术处理PDF文件。该数据集主要用于评估LLMs在天文研究中的事实回忆和基于当前天文共识的广泛推理能力。
The AstroLLaMA-2-70B dataset was developed by a research team from the University of Illinois Urbana-Champaign, with the goal of evaluating the performance of astronomy-focused large language models (LLMs). This dataset contains 4,425 multiple-choice questions (MCQs) spanning a wide range of astronomical topics and concepts. The dataset construction process includes extracting abstracts, introductions, and conclusions from the astrophysics section of arXiv, as well as processing PDF files via optical character recognition (OCR) technology. This dataset is primarily used to evaluate the factual recall and general reasoning capabilities of LLMs in astronomical research based on current astronomical consensuses.

- 1AstroMLab 2: AstroLLaMA-2-70B Model and Benchmarking Specialised LLMs for Astronomy伊利诺伊大学厄巴纳-香槟分校 · 2024年



